Skip to main content

Table 3 Mean precision (P), recall (R), and F1-measure (F) 10-fold cross validation results on training data for the most frequent properties per scene

From: Personality-dependent content selection in natural language generation systems

 Majority class baselinePersonality-dependent CS
ScenePropertyInst.+/ −PRFPRF
01red-sandals46/1050.701.000.820.820.960.88
01brown-dress37/1140.761.000.860.860.990.92
01long-dress26/1250.831.000.910.840.980.90
02black-canvas74/770.511.000.680.810.990.89
02white-bucket65/860.571.000.730.760.980.86
02large-bucket30/1210.801.000.890.861.000.92
03fat-lips89/620.561.000.710.770.990.87
03white-bedspread59/920.621.000.760.740.990.85
03red-details56/950.541.000.780.750.990.85
04black-person75/760.521.000.690.840.970.90
04raised-arms35/1160.621.000.760.731.000.84
05sitting-people38/1130.751.000.860.860.960.81
05green-coat38/1130.751.000.860.771.000.87
05white-hair28/1230.811.000.900.771.000.87
06pink-bedspread79/720.521.000.680.790.990.88
06blue-blanket74/770.511.000.680.830.970.89
06blue-coat67/840.561.000.710.770.990.87
08black-man111/400.721.000.840.860.960.91
08blue-cap35/1160.771.000.870.721.000.84
08barbed-wire25/1260.831.000.910.851.000.92
09yellow-shorts51/1000.541.000.700.750.980.85
09plaid-shorts45/1060.701.000.820.800.990.89
09blue-shirt37/1140.751.000.860.770.970.86
10dry-grassland40/1110.741.000.850.830.990.90
10elderly-woman30/1210.801.000.890.830.990.90
10red-pants30/1210.801.000.890.751.000.86
 Mean51/1000.681.000.800.800.990.88
  1. Best F1 results are highlighted